Leveraging Molecular Data for Drug Discovery
summary
Proposes a rank-based count statistic with derived asymptotic properties to detect drug–disease expression reversals and a gene-level score to flag potential drug targets; validates on simulations and real drug‑discovery datasets where the measures match or outperform existing metrics. Develops a deep matrix factorization denoising pipeline with residual thresholding applied to five large pharmacogenomic matrices, increases drug‑wise agreement across datasets and sharpens separation of mutated versus non‑mutated cell lines, and runs with scalable gradient‑descent optimization and early stopping.
AI-written summary · about
chapters (4)
related talks
30mMachine Learning: A New Approach to Drug Discovery
33mNetwork Data Integration Enables Precision Medicine
36mFormatting Biological Big Data to Enable (Personalized) Systems Pharmacology
53mThe Algorithm for Precision Medicine
35mFrom cells to drug responses - machine learning in cancer research
33mUnderstanding Molecular Complexity for Precision Medicine
44mNetwork Propagation as a Tool for Deciphering Disease Mechanisms
29mMachine Learning Techniques for Omics Data Analysis